SenseNova-SI-1.1-Qwen3-VL-8B-GGUF

SenseNova-SI-1.1-Qwen3-VL-8B-GGUF

SenseNova-SI-1.1-Qwen3-VL-8B from sensenova is an 8B-parameter multimodal vision-language model fine-tuned from Qwen/Qwen3-VL-8B-Instruct as part of the SenseNova-SI family, specifically scaled on 8M diverse spatial intelligence samples (SenseNova-SI-8M) to address deficiencies in spatial reasoning, achieving state-of-the-art open-source performance on benchmarks like 64.8% VSI, 38.1% MMSI, 73.8% MindCube-Tiny, 51.2% ViewSpatial, and 49.6% SITE—surpassing peers like InternVL3-8B, VST-7B-SFT, and approaching proprietary models such as Gemini-2.5-Pro. It excels in robust spatial capabilities across VSI-Bench, MMSI, MindCube, ViewSpatial, SITE, BLINK, 3DSRBench, and EmbSpatial-Bench while retaining strong general multimodal understanding (84.9% MMBench-En), demonstrating emergent generalization from data scaling, reduced overfitting risks, and preliminary spatial chain-of-thought reasoning for downstream applications. Released publicly to advance spatial intelligence research (with future integration into larger in-house models), it supports evaluation via EASI framework and evaluation on eight recent benchmarks, positioning it as a high-performing, compatible foundation for further spatial AI development.

SenseNova-SI-1.1-Qwen3-VL-8B [GGUF]

Quants Usage

(sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants)

Here is a handy graph by ikawrakow comparing some lower-quality quant types (lower is better):